Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
NOVATEUR PUBLICATIONS
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
A FRAMEWORK AND ALGORITHMS FOR ENERGY EFFICIENT CONTAINER
CONSOLIDATION IN CLOUD DATA CENTERS
Mr. Mahabaleshwar M. Mundashi
M. Tech Student
Dept. of Computer science and Engineering.
Rajarambapu Institute of Technology, Sakhrale, India
mmmundashi555@gmail.com
Abstract:
Now a day’s utilization of power in cloud data centers is very
large. Minimization of power consumption in data centers is major
challenge. Cloud computing uses a network of remote servers
which are hosted on the Internet to process, manage, and store
data rather than a local server. Containers in cloud is nothing but
an image which is a lightweight, software package which executes
independently which hold everything you need to run: code,
system libraries, system tools, settings. Using container
consolidation, containers are migrated to number of hosts. In
Container overbooking, allocate some percentage of workload to
any resources. In this scenario hosts having CPU utilization and
power consumption less than 70% or more than 80% are
considered as underloaded or overloaded. In this paper, focus is
on the performance of two algorithms named container
overbooking and initial placement of containers and compare the
results with scheduling algorithms named FirstFit, MostFull,
Random and LeastFull. We will combine the both algorithm and
create the algorithm which improves the results with reduced the
power consumption in data center and resource utilization.
Keywords: Cloud, energy consumption, cloud data centers,
Container consolidation, overbooking
I. INTRODUCTION
Cloud computing is also known as “the cloud”, it uses
remote servers which are hosted on internet for storage,
manage, and processing of data that are delivered on-demand
on a pay-for-use basis. Users work remotely using cloud
computing. The resources retrieved from the cloud are through
web based tool and applications. Cloud computing gives
several benefits for business users and clients.
 Self-service provisioning: End users can set up and launch
application services in cloud environment and use those
resources on demand for any type of workload. This put an
end to the conventional requirement for IT administrators to
supply and manage compute resources.
 Elasticity: There will be increase in computing needs and
also decrease in the needs therefore companies can scale up
and down as per demand. It removes the demand of huge
expenditure in local infrastructure which may or may not
remain active.
 Pay for use: Users use those resources for which they are
paid and workloads they use. [4]
The container is nothing but an image which is a
lightweight, software package which executes independently
126 | P a g e
April, 13th, 2018
Prof. S. R. Poojara
Assistant Professor
Dept. of Computer science and Engineering.
Rajarambapu Institute of Technology, Sakhrale, India
shivananda.poojara@ritindia.edu
which hold everything you need to run: code, system
libraries, system tools, settings. The containerized software is
available for Linux and Windows based apps and runs the same
regardless of the environment. Containers running on the host
machine share the kernel of host’s operating system; it starts
immediately and uses less compute resources and RAM.
Images are created from the layers of the file system and share
common files. This minimizes disk usage and image uploading
is much faster. [1][2]
Consolidation strategy combines the containers which are
marked as overloaded. This technique is used for minimization
of energy utilization in cloud data centers, containers are
migrated on number of hosts which are having smaller support
of ContainerCloudSim framework.
This paper focuses on container overbooking and container
placement policy algorithms. Container overbooking algorithm
first check whether the container is in overbooked or what then
if it is overbooking condition then that container needs to
migrated to another host. For that it will moved to container
migration list. In another algorithm container placement policy
uses three algorithms FirstFit, MostFull and Random algorithm.
Based on the results and performance evaluation of above
designed algorithms we are trying to design such an algorithm
which improves the results of algorithm and reduces the energy
consumption in data centers with resource utilization.
Work is carried out in following format; Section 2
elaborates previous work done and technique used for that work
by researchers in CloudSim. Section 3 gives brief introduction
about ContainerCloudSim, need of simulation and its
taxonomy. In section 4 details working about proposed research
regarding the Container overbooking, consolidation, container
placement policy in Cloud Data Centers by using
ContainerCloudSim framework and algorithms and also
mathematical model for power consumption in data centers are
stated. Section 5 represents discussion about experiment setup
and results. Chapter 6 outlines the conclusions of proposed
study.
II. RELATEDWORK
In paper [1], Claus Pahl et al. give an introduction to
Virtualization and need for the Containerization. It has been
discussed that to manage and orchestrate applications PaaS
cloud service can use containers. Overall the experiment shows
about container containerization technologies and how
applications can be deployed and run on PaaS platform.
 Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
In paper [2], Sareh Fotuhi et al. author gives the brief
information about ContainerCloudSim. Container is having
various advantages over the use of virtual machine.
ContainerCloudSim is a framework or tool which facilitates
containerized cloud computing environments simulation and
modeling support for cloud data center. In this paper author had
developed simulation architecture and apply various algorithms
on containerized cloud data center and executed it as an
extension of CloudSim.
In paper [3], Piraghaj et al. carried out CloudSim
architecture and various algorithms are applied using CloudSim
for minimization of energy utilization in cloud data center with
required Quality of Service (QoS). This paper carried out the
whole information and related algorithms for simulation of
containers, container consolidation techniques. The objective of
authors is to address the betterment in the utilization of
resource on server and virtual machine.
In paper [5], Piraghaj et al. told significant problems that
cloud providers face for reducing total energy utilization in
cloud data centers by using various algorithms and CloudSim
tool or framework. The main focus of author is improving
energy efficiency of servers by using the CloudSim framework.
This paper introduces about CaaS Service and also impact of
container overbooking, container migration etc.
In paper [6], Amir Vahid Dastjerdi et al. states
classification of various resource management techniques for
energy efficiency. Author says that energy utilization in cloud
data centers not only affects the overall cost but also it
increases the environmental problem as carbon dioxide
emission.
In paper [7], Awada et al. author proposed that energy
consumption becoming key issue for the operation and
maintenance of cloud datacenters; cloud computing providers
are becoming deeply concerned. In this paper, author presents
formulation and solution for Green Cloud Environments to
reduce environmental impact and energy consumption by
examining static and dynamic portions of cloud components
under new models.
In paper [9], M. Khoshkholghi et. al. states that cloud data
centers can utilize huge amount of energy which contributes to
high operational cost and emission of carbon dioxide. To
reduce the resource utilization and reduction of energy
consumption author proposes a technique dynamic
consolidation of VM using migration but VM consolidation
may decreases the performance. In this paper different
algorithms are proposed for consolidation of VM dynamically.
In paper [10], Buyya Rajkumar et al. proposed the
CloudSim framework for simulation of data centers in cloud
computing environments. CloudSim is used to simplify the
application deployment, how resources are allocated and for
scheduling algorithms. CloudSim supports one or more virtual
machines for modeling and simulation.
In paper [11], Mehiar Dabbagh, Bechir Hamdaoui et al.
proposed a framework that is integrated energy-aware resource
provisioning for energy consumption in cloud data centers. It
provides physical machines to cloud data centers for providing
services to their clients. Framework contains three main
127 | P a g e
NOVATEUR PUBLICATIONS
April, 13th, 2018
components: power management, clustering and workload
prediction.
In paper [12], Wang Long, Lan Yuqing and Xia Qingxin et
al states that simulation will reduce the effort to configure
testing environment and avoids the spending time. In this paper
author uses CloudSim framework for simulation provides
simulation, power to manage services and to model the cloud
infrastructure. It also states VM allocation policy, VM selection
policy.
In paper [13], Anton Beloglazov et al. says that cloud data
centers consume huge amount of electrical power which may
results in high operational cost and carbon dioxide emission.
The objective of author is minimize energy consumption by
migrating the VMs on another host and turn off idle nodes in
cloud data center by providing required quality of service to the
users.
In paper [14], Kantarci, Burak, H. Mouftah covers the
energy efficiency in cloud data center which contains many
aspects such as storage and network related power
consumption. For energy efficiency in cloud computing author
proposes various aspects like processing, storage and power
consumption network etc.
In paper [15], Berl, Andreas, et al. said that efficiency of
energy is essential for future information and communication
technologies. In this paper author focuses on efficient energy
scheduling in multiprocessor and grid system, minimization of
energy in clusters of servers and also impact of energy saving
strategies for management of integrated system.
In paper [16], Qouneh, Amer, Ming Liu, and Tao Li., states
that energy utilization and performance are two main
challenges in cloud computing data centers. To decrease energy
consumption and maintain performance author uses two
techniques that is, take power from co-located multithreaded
virtual machines (VMs) and distribute it to VMs and
compensate multi-threaded VMs and re-boost their
performance.
In paper [17], Park, KyoungSoo, and Vivek S. Pai., says
that CoMon is an evolving, mostly-scalable monitoring system
for PlanetLab. In this paper author designs what kinds of data
are gathered in PlanetLab.
In paper [18], Kulseitova, Aruzhan, and Ang Tan Fong.,
proposed that there is need for applying some energy efficiency
techniques in cloud data centers. Thus author present a recent
research in cloud data centers and gives taxonomy of work.
III. Definition and Model
A. ContainerCloudSim:
ContainerCloudSim is an extension of CloudSim simulator
framework. ContainerCloudSim is designed for migration of
containers from source host to destination host based on the
power consumption in data centers. ContainerCloudSim uses
techniques like container overbooking, container placement,
scheduling and consolidation of containers. Container as a
Service’s cloud data center which contains virtual machines,
hosts, containers and applications with their workloads.
ContainerCloudSim is divided into two parts that is simulated
elements and simulated services for implementation. Aim of
 Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
this technique is to provide energy efficient cloud data centers
by migrating container from source host to destination host by
applying some algorithms. [2][8]
B. Need of simulation in Cloud Datacenter:
Simulation is widely used technology used frequently to
examine complex problems, most simulation-based
experiments consider simplified modeling of cloud computing
and application environments. CloudSim is a library for
simulating cloud scenarios. It provides essential classes for
describing data centers, compute resources, virtual machines,
applications, users, and management strategies for various parts
of the system, such as planning and provisioning.
Simulation provides a safe way to test and explore
different scenarios. Simulation is used for improvement in
performance and energy efficient cloud data center. It is used to
know in which ways a part could fail and what loads it can
oppose which is helpful for designers and engineers to
understand conditions. Simulation is used in many conditions
such as, simulation for performance optimization, testing,
scientific experiments etc. The main advantage of simulation is
it provides users feedback practically while designing real-
world systems.
C. Taxonomy of ContainerCloudSim
Fig. 3.1 System Architecture and Processes
Figure 3.1 [5] represents the ContainerCloudSim system
architecture and processes. The main components of
ContainerCloudSim are
 Host Status Module
Host Overload/ Underload Detector – It is widely used
mechanism. In this module host is established as underloaded
128 | P a g e
NOVATEUR PUBLICATIONS
April, 13th, 2018
or overloaded. If the host is established underloaded then host
detector sends host ID and containers ID which are running on
that host to consolidation module. And if it established
overloaded then detector send request to the Container selector
component for activating component.
Container Selector – This component is activated when host in
cloud data center is established as over-loaded. This process of
selection of containers is goes on unless host status is available
in between underloaded and overloaded condition.
Container Migration List (CML) – This component stores the
containers information which are selected by Container
Selector component and then submit that information to
consolidation module.
 Consolidation Module
Over-loaded Host list – Task of this component is to store
over-loaded host.
Over-loaded Destination Selector – It takes input as
overloaded host list, list of containers migration and active host
list. Then it migrates the over-loaded host to destination host.
Destination list – This component contains data having
container ID with host and VM ID both are received from the
above component.
VM Creator – The most important consideration of this
component is to create large number of VM’s in underloaded
host and assigns containers to that VM’s. If any container is
remaining, then it selects host randomly from the idle host and
creates VM on that host.
Underloaded host list – The status of the host that are found to
be under-loaded are stored in this component.
Underloaded Destination Selector – The mechanism of this
task is to find the best destination host for containers from
underloaded hosts by using host selection algorithm. This
component sends the containers to the destination which is
decided by overloaded destination selector to VM-Host
Migration Component along with host ID.
VM-Host Migration Manager – This is used for triggering the
migration of saved containers ID together with selected
destination.
Under-loaded Host Deactivator – After migrating all the
containers this component switches off under-loaded hosts.
IV. EXPERIMENTAL SETUP
To deploy ContainerCloudSim on Eclipse IDE we need to
download Simulator which is .zip file of CloudSim-4.0 from
GitHub and also download eclipse. Deploy the CloudSim-4.0
by extracting the .zip file and import it in Eclipse.
1. Container overbooking
Overbooking manages the trade-off between maximizing
resource utilization and minimizing performance degradation
and SLA violation. [2] By allocating some percentage of
workload to any resources ContainerCloudSim is capable of
container overbooking. In this scenario hosts having CPU
 Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
NOVATEUR PUBLICATIONS
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
April, 13th, 2018
utilization and power consumption less than 70% or more than The consolidation algorithms analyze the energy efficiency
80% are considered as overloaded or underloaded. based on the data center energy consumption obtained from
Equation 1.
Algorithm: Container Overbooking
Input: list of container in server (SCL) b) SLA Metric –
Output: selected container
Step 1 begin with input while SLA = (4)
Step 2 check for host overloaded
If host status marked as overloaded then go to step 3
Else check host status
either overloaded or underloaded A Service-Level Agreement is an agreement between two or
Step 3 Apply container selection algorithm more parties, where one is the customer and the others are
to get container from container selection list(SCL) service providers. This can be a legally binding formal or an
Then add selected container to the list informal "contract". Service-Level Agreement is a part of cloud
computing to ensure maximum availability of services for
2. Container consolidation customer. With violation of SLA, the provider has to pay
This technique is used for reducing energy consumption in penalties.
data centers. Using consolidation technique, containers are c) Problem Formulation – For minimizing power utilization
migrated to number of hosts which are smaller with support of of data center with M containers, N VM’s and K servers, we
ContainerCloudSim framework. [5] Migration of containers is formulate the problem as follows:
necessary because host is either overloaded or under-loaded. min(Pdc(t) = (5)
Here two algorithms are used “Most Correlated” and “Max
Usuage” for container selection. Max usage algorithm select the
containers which has most CPU utilization while most (6)
correlated algorithm chooses containers CPU workload is most
correlated with host. [2] ,
Finally at last stage three bin packing algorithms for host
(7)
selection are applied: FirstFit Host Selection, Random Host
Selection and Least Full Host Selection.

3. Container placement policy

Here we study different container to VM mapping algorithms
to see how ContainerCloudSim is used to carry out systematic , (8)
inquiry of algorithms on container placement for total power
consumption, number of container migrations in data center and ,
resulting SLA violations. FirstFit, MostFull and Random are
three different placement policies are evaluated. [5] , (9)
Symbols Description
 Mathematical model for power consumption in Cloud Pdc(t) Power Consumption of the data center at time t
Datacenter [5] Pi(t) Power Consumption of server i at time t
Ns Number of servers
a) Data Center Power Model - Energy efficiency of the data PiIdle Idle power consumption of server i
center at time t is calculated as: Pimax Maximum power consumption of server i
Ui,t CPU utilization percentage of server i at time t
Pdc(t) = (1) Nvm Number of vms
Nc Number of containers
For each server i, utilization of CPU (Ui,t) is equal to Uc(k,j,i)(t) CPU utilization of container k on (VM j, server i)
at time t
Nv Number of SLA Violations
tp The time t at which the violation p happened
Vm j i VM j on server i
CPUr (vm CPU amount requested by VM j on server i at
and the power consumption of the server is estimated using j i, tp) time tp
Equation 2. CPUa (vm CPU amount allocated to VM j at time t p
j i, tp)
Pi (t) = Piidle+ (Pi max- Piidle)* Ui,t NVM> 0 (2) S(i,r) Server i Capacity for resource r
Uvm j,i (t) CPU utilization of VM j on server i at time t
0 NVM= 0 (3) vm(j,i,r) The capacity of resource r of VM j on server i
C(k,j,i,r) The resource r capacity of container k on (VM j,
server i)
Table: Symbol table [5]

129 | P a g e
 Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
V. RESULT AND DISCUSSION
We evaluate two algorithms with two methods. We are
used ContainerCloudSim to conduct simulation with different
parameter such as overbooking factor, container migration rate,
execution time, container placement etc.,
A. Container Overbooking
By allocating some percentage of workload to any
resources ContainerCloudSim is capable of container
overbooking. In this scenario hosts having CPU utilization and
power consumption less than 70% or more than 80% are
considered as overloaded or underloaded.
NOVATEUR PUBLICATIONS
April, 13th, 2018
Based on the figure 3 and figure 4 we can say that FirstFit
algorithm for container overbooking performs better as
compare to MostFull algorithm. The output of the simulation
shows that number of successfully allocated containers
decreases as percentage of energy consumption increases. The
volatility of workload is the key factor that affects the
percentage of value. FirstFit algorithm results in less number of
migration and energy consumption and thus should be preferred
policy to be utilized as the goal of provider is to reduce energy
consumption.

B. Container Placement Policy

Here we study different container to VM mapping
algorithms to see how ContainerCloudSim is used to carry out
systematic inquiry of algorithms on container placement for
total power consumption, number of container migrations in
data center and resulting SLA violations. FirstFit, MostFull and
Random are three different placement policies are evaluated.

 Scheduling Algorithms
FirstFit – In the first fit, containers are allocated to virtual
machines which are first sufficient from the top of Main
Memory. In FirstFit algorithm, available host is opened when it Fig. 5.2 Container Overbooking (MostFull)
receives the first container. When all the containers in a host
are removed from source host, the host is closed. FirstFit
algorithm results in less number of migrations and energy
consumption.

MostFull – In the most full, containers are allocated to virtual

machines which are most correlated with host. MostFull
algorithm packs container on most full virtual machine in terms
of CPU utilization, which results in higher migration rate.

Random – It typically uses random bits as an input in the hope

of achieving good performance in the average case over all
possible choices of random bits. In random algorithm,
Fig. 5.3 Container Initial Placement (FirstFit)
containers are allocated to virtual machines randomly.

Fig. 5.1 Container Overbooking (FirstFit)

Fig. 5.4 Container Initial Placement (MostFull)
130 | P a g e
 Proceedings of 4th RIT Post Graduates Conference (RIT PG Con-18)
JournalNX- A Multidisciplinary Peer Reviewed Journal (ISSN No:2581-4230)
Figure 5.3 and 5.4 states that how containers which are in
overbooking condition are initially placed onto another host.
Based on the figure we can say that container migration rate for
FirstFit algorithm is less than MostFull algorithm. The most
full container placement algorithm packs containers on most
full most full virtual machine in terms of CPU utilization,
results in higher container migration rate. Consequently it
results in higher violation and energy consumption.
VI. CONCLUSION
ContainerCloudSim is a framework which used to enquire
the execution of container placement algorithms on number of
container migrations, data center total power consumption and
SLA violations and for implementing energy efficiency
techniques. In this work we design the container overbooking
for FirstFit and MostFull container allocation policy and
Container initial placement policy for FirstFit and MostFull.
The work suggests that the designed strategy is helpful for
improving efficiency of energy in cloud data centers. This
algorithm checks the hosts which are over-loaded and store
them into container migration list. Then it checks for hosts
which are free. In the next work we will use container
consolidation technique with the combination of container
overbooking and container placement policies for improvement
in energy efficiency at cloud data centers.
NOVATEUR PUBLICATIONS
April, 13th, 2018
[12] Long, Wang, Lan Yuqing, and Xia Qingxin. "Using cloudsim to model
and simulate cloud computing environment." Computational Intelligence and
Security (CIS), 9th International Conference on. IEEE, 2013.
[13] Beloglazov, Anton, and Rajkumar Buyya. "Energy efficient allocation of
virtual machines in cloud data centers." Cluster, Cloud and Grid Computing
(CCGrid), 10th IEEE/ACM International Conference on. IEEE, 2010.
[14] Kantarci, Burak, and H. Mouftah. "Energy-Efficiency in Cloud Data
Centers." Comm. Infrastructures for Cloud Computing 241-263, 2013.
[15] Berl, Andreas, et al. "Energy-efficient cloud computing." The computer
journal 53.7, 1045-1051, 2010
[16] Qouneh, Amer, Ming Liu, and Tao Li. "Optimization of resource
allocation and energy efficiency in heterogeneous cloud data centers." Parallel
Processing (ICPP), 44th International Conference on. IEEE, 2015.
[17] Park, KyoungSoo, and Vivek S. Pai. "CoMon: a mostly-scalable
monitoring system for PlanetLab." ACM SIGOPS Operating Systems Review
40.1, 65-74 2006.
[18] Kulseitova, Aruzhan, and Ang Tan Fong. "A survey of energy-efficient
techniques in cloud data centers." ICT for Smart Society (ICISS), International
Conference on. IEEE, 2013.

REFERENCES
[1] C. Pahl, "Containerization and the PaaS Cloud", IEEE Cloud Computing,
vol. 2, no. 3, pp. 24-31, 2015.
[2] S. Piraghaj, A. Dastjerdi, R. Calheiros and R. Buyya, "ContainerCloudSim:
An environment for modeling and simulation of containers in cloud data
centers", Software: Practice and Experience, 2016.
[3] Piraghaj, Sareh Fotuhi. "Energy-Efficient Management of Resources in
Container-based Clouds." (2016).
[4] Rajkumar Buyya, Christian Vecchiola, and S. ThamaraiSelvi. "Mastering
Cloud Computing." (2013): 1-1.
[5] Piraghaj, Sareh Fotuhi, et al. "A framework and algorithm for energy
efficient container consolidation in cloud data centers." Data Science and Data
Intensive Systems (DSDIS), International Conference on. IEEE, 2015.
[6] Amir Vahid Dastjerdi, Rodrigo N.Calheiros, and Rajkumar Buyya, A
Survey and Taxonomy of Energy Efficient Resource Management Techniques
in Platform as a Service Cloud, Handbook of Research on End-to-End Cloud
Computing Architecture Design Book, J.Chen, Y.Zhang, and R.Gottschalk
(eds), IGI Global, Pages. 410 - 454. Web. 16 Oct. 2016. doi: 10.4018/978-1-
5225-0759-8.ch017, Hershey, PA, USA, 2017.
[7] Awada, Uchechukwu, Keqiu Li, and Yanming Shen. "Energy consumption
in cloud computing data centers." International Journal of Cloud Computing
and services science 3.3 (2014).
[8] www.cloudbus.org/cloudsim
[9] M. Khoshkholghi, M. Derahman, A. Abdullah, S. Subramaniam and M.
Othman, "Energy-Efficient Algorithms for Dynamic Virtual Machine
Consolidation in Cloud Data Centers", IEEE Access, vol. 5, pp. 10709-10722,
2017.
[10] Buyya, Rajkumar, Rajiv Ranjan, and Rodrigo N. Calheiros. "Modeling and
simulation of scalable Cloud computing environments and the CloudSim
toolkit: Challenges and opportunities." High Performance Computing
Simulation, HPCS' International Conference on. IEEE, 2009.
[11] M. Dabbagh, B. Hamdaoui, M. Guizani and A. Rayes, "Energy-Efficient
Resource Allocation and Provisioning Framework for Cloud Data Centers",
IEEE Transactions on Network and Service Management, vol. 12, no. 3, pp.
377-391, 2015.

131 | P a g e